Intel X79 Info Leaked - Insights into I/O on Sandy Bridge Mac Pros?

[VirtualRain]

This leaked information on the X79 chipset should provide some insights into what to expect with a Sandy Bridge refresh of the Mac Pro either late this year or early next...

Chinese site it.com.cn has got their hands on newest roadmap of the Intel X79 platform. According to the site and roadmap, X79 will be the successor to the three year old then X58 platform in Q4 2011. As many have already known, the new processors will be called the "Sandy Bridge E" series and will be seen with 2011 land grids, therefore compatible with X79 motherboards of socket LGA2011.

Major improvements were to be seen in X79. Rumours were certain that the new platform and CPUs will beat the current flagship Intel Core i7-990X hands down. Revealing even more information from Intel's roadmap were some features to be seen in X79 as shown below.

The new processors will only be available in four cores or six cores and expected to have a larger cache memory than current processors. X79 will support quad channel DDR3 memory and native support of memory speeds up to 2666MHz. These information are hard to believe but possible.

Those currently on the P67 platform will know that a internal clock generator will largely limit BCLK overclocking. Sad to say, based on the source, the clock generator of Sandy Bridge E series will again be within the CPU. However, these are just rumours and should not be taken to heart.

PCI interface will totally be removed from the X79 chipset while on the other hand, up to 10 native SATA 6Gbps interface will be available with a total number of possible SATA ports at 14. In order to have more bandwidth for the DMI Interface(or whatever name Intel wants to name it in the new platform), an additional PCI-E 2.0 x4 can directed to it making a total bandwidth of 8GB/s. On a side note, USB3.0 will not be there and only enabled via an external controller such as one from NEC. SAS capabilities with 8 of the SATA ports is a good thing to be seen on the new platform as well.

A high end platform can't be complete without sufficient graphics bandwidth and on the X79, there will be at least 32 PCI-Express 2.0 lanes. So similar to X58, graphics can run at dual x16 or quad x8.

With all the advantages in X79, many who were looking forward to technology such as PCI-E 3.0 as well as Intel ThunderBolt were saddened with the fact that they will not be featured.

The only thing skeptical about X79 is its overclocking capabilities. The overclocking community would not want to know that X79 cannot scale well with sub zero temperatures. With disappointing P67 on this aspect, we can only wait until the actual release date arrives and try it out ourselves.

Source: http://vr-zone.com/articles/intel-x79-roadmap-leaked/11762.html#ixzz1IICW6Bmh

The number of SATA(3) ports is interesting, if not hard to believe. SAS support is also interesting although less relevant to enthusiasts these days in light of SSDs. However, these added ports could be the catalyst for a chassis redesign to accommodate more internal drives.

If the RAID capabilities of this chipset are true, it will definitely put a damper on the aftermarket RAID card market place. Let's just hope that Apple supports using the chipset RAID features (something they haven't in the past).

Notable omissions are Thunderbolt, USB3, and PCIe3. However, I'm sure the next Mac Pro will include Thunderbolt via an extra chip.

 

[Transporteur]

8GB/s bandwidth for the SATA ports? Oy boy! SSD RAID here we go!
The current systems have 660MB/s. How pathetic.

I'm so glad I didn't jump for the 2010 Mac Pro. Those Sandy Bridge things will probably be a massive upgrade.

 

[nanofrog]

They've combined the ICH into the chipset, so the ability to support RAID functions isn't a surprise at all really (think R versions, = RAID). But it's nowhere near as fast or robust (think recovery features), as a proper hardware RAID card.

As per PCIe 3.0, I have to presume that they don't expect any such devices in the near future (we've not even maxed out PCIe 2.0 yet anyway), and it saves on costs as well, so that's not really a suprise (saddening, but not a surprise).

TB requires a separate chip, and Intel's holding up on USB 3.0 to give it a chance at adoption, so no surprise there either.

The SATA ports will be nice. As per SAS, that's a nice addition for the enterprise market, which is what this chip is actually aimed at (head on anyway).

 

[VirtualRain]

They've combined the ICH into the chipset, so the ability to support RAID functions isn't a surprise at all really (think R versions, = RAID). But it's nowhere near as fast or robust (think recovery features), as a proper hardware RAID card.

What do you make of the "e" for "enterprise" appended to the RST in these slides? Is that new? I wonder what it implies? Perhaps some enhanced RAID 5 support with dedicated logic for parity calculations or even recovery operations? Any thoughts?

 

[nanofrog]

What do you make of the "e" for "enterprise" appended to the RST in these slides? Is that new? I wonder what it implies? Perhaps some enhanced RAID 5 support with dedicated logic for parity calculations or even recovery operations? Any thoughts?

Though e for enterprise makes sense to me (LGA2011 will be enterprise parts), the second (X79 slide image, not the roadmap), E is stated to mean Enthusiast.

But I also expect this chipset to be used with the LGA1356 parts (Socket B2) due out as well (serve as both performance desktops and workstations; separation is ECC Enabled or Disabled as it has been since Nehalem). They can cut costs this way (one chip that supports 2x sockets). BTW, they could call the consumer versions either different numbers under i7 (would add further confusion IMO, but quite possible), or use i8 for them, with the workstation versions remaining under the Xeon moniker.

Assuming my thinking matches theirs, calling it an Enthusiast part is valid, as it will be used with those parts as well.

 

[VirtualRain]

Though e for enterprise makes sense to me (LGA2011 will be enterprise parts), the second (X79 slide image, not the roadmap), E is stated to mean Enthusiast.

Perhaps you missed what I was saying... For the first time that I'm aware, Intel has dubbed their chipset RAID technology as RSTe (Rapid Storage Tchnology Enterprise) where the "e" and "enterprise" are new additions to this moniker. See slide 2. It makes me wonder what new chipset RAID capabilities might have warranted the new branding.

 

[Hellhammer]

BTW, they could call the consumer versions either different numbers under i7 (would add further confusion IMO, but quite possible), or use i8 for them, with the workstation versions remaining under the Xeon moniker.

But isn't that what Intel aims at with their naming systems?

As usual, nanofrog pretty much said everything worth saying. It's a bit weird though that according to rumors, Panther Point will support PCIe 3.0 and USB 3.0 and it is meant for consumers. Where are the good ol' times that high-end parts got these new technologies first?

 

[nanofrog]

But isn't that what Intel aims at with their naming systems?

Certainly seems that way, doesn't it.

Actually if it made sense, a lot more users may cease buying more CPU than they need, thus make a noticable dent in their bottom line. So they keep it confusion on purpose.

It's a bit weird though that according to rumors, Panther Point will support PCIe 3.0 and USB 3.0 and it is meant for consumers. Where are the good ol' times that high-end parts got these new technologies first?

It's due to release later (gives them more time to get the design sorted properly), and may also coincide with the spin down of another product using a smaller die process.

Both of these have significant cost implications (time = money, and the enterprise stuff is more complex, which needs more time to get right). Combining parts (i.e. chipset + separate ICH into a single part) and producing them on smaller die processes make production cheaper for Intel.

Since the consumer parts aren't as complex (read smaller design due to lower complexity), they can be produced faster, and may not need to be on the smallest die size possible to get sufficient parts per wafer to be profitable.